Cloud-based computing is rapidly gaining considerable interest in both consumer and business markets, as it provides remote computer resources on demand, such that the user no longer has to rely on investing in hardware to host the services, e.g., applications, required by the user. Such services can range from computer programs such as e-mail clients, word processors and the like to commercial services such as information retrieval services, web-based trading services, e.g., e-shops and the like.
A typical cloud-based computing architecture includes a cloud management server that manages access to the resources that form part of the cloud. A user typically contacts the cloud management server when requesting a particular service, which triggers the cloud management to locate a resource in the cloud that can host the requested service, e.g., by broadcasting a service request to all cloud servers. Upon locating a cloud server available to host the required service, the available cloud server will be assigned the task of hosting the requested service for the user, upon which the cloud server typically will launch an instance, e.g., a virtual machine (VM) for running the requested service. The cloud instances are usually managed in such a way that it allows users to create, update and delete the instance, either manually or by automatically by an application programming interface (API), such as Representational State Transfer (REST) or a language-specific API.
In such a cloud-based computing architecture, many VMs may be launched over a period of time. This can cause problems if the VMs are not terminated upon completion of the service, as this prevents computer resources being freed up, thus unnecessarily reducing the available capacity of the cloud. This problem is particularly prevalent in architectures in which the user is expected to terminate the VM upon delivery of the requested service, as many users tend to forget that they are responsible for such terminations, which can lead to many idle VMs being present in the architecture.
In some cases, the problem is addressed by the cloud management system inserting a self-management module into the VM, which can terminate the VM if a predefined condition is met, such as the occurrence of an error in the VM, the end of a predefined lifespan or completion of a specified duration of the VM, upon reaching a user credit limit for the use of the VM or upon an inactivity in the resources consumed or utilized by the VM. However, measuring inactivity by resource consumption is not a particularly accurate metric for determining VM inactivity, such that a risk exists that a VM may be erroneously terminated. This is of course highly unwanted, as it severely reduces the confidence the user has in the cloud computing architecture.